Anthropic has announced that, beginning November 12, it will prohibit "sustained and needless abusive or cruel behavior" toward its AI models. The rule is framed as a usage policy violation. The company says the measure promotes responsible use and aims to prevent harm that could arise from such interactions.
Aaron Levie, the chief executive officer of Box, offered a supportive view of the policy. He said, "Even if you don't believe AI is conscious — I don't — it stands to reason that you don't want future models trained on endless content of humans being rude to models." Levie’s comment suggests that, regardless of whether one attributes consciousness to the systems, limiting exposure to hostile language helps avoid incorporating negative human behavior into the training data. His stance is presented as a precautionary measure intended to protect the quality of future model outputs.
Thou shall not bully Claude, Anthropic will ban you if you do https://t.co/v0EdUHcudL
Michael Shellenberger responded critically, calling the policy "anthropomorphizing machines."
Michael Shellenberger responded critically, calling the policy "anthropomorphizing machines.
Bill Gurley raised a different concern, asking whether the policy implies that Anthropic is using customer prompts as part of its training data. He phrased it as a question: "Strikes me this implies the company is certainly using customer's prompts as part of the training data?"
The announcement has set off a conversation among technology leaders about how to govern interactions with AI systems. While some see the rule as a sensible step to curb harmful training influences, others warn against attributing human traits to algorithms, and a third line of questioning focuses on what the policy might reveal about data practices. The debate remains open, with each commentator interpreting the same policy through a different lens about the relationship between people and the AI models they use.